| Operators |
select_characters — Selects characters from a given region.
select_characters(Region : RegionCharacters : DotPrint, StrokeWidth, CharWidth, CharHeight, Punctuation, DiacriticMarks, PartitionMethod, PartitionLines, FragmentDistance, ConnectFragments, ClutterSizeMax, StopAfter : )
select_characters selects from a given Region the areas which might be characters and returns them in RegionCharacters. This is done by using features like StrokeWidth, DotPrint, the size of the characters and some more. The given Region should be united, else every Region is processed separately. Thus do not call connection before calling select_characters, because then fragments or dots would not be connected to a character. If you have more than one region with text, you can of course handle them without merging them. The Region for select_characters typically comes from segment_characters but also any other segmentation operators can be used.
The process of the selection can be partitioned into four parts. All steps are influenced by the parameters StrokeWidth, CharHeight, and CharWidth. If you loose small objects like dots, adapt the minimum CharWidth and the minimum CharHeight. But some parameters affect the result of a certain step in particular. A closer description follows below. With the parameter StopAfter you can terminate after a specified step.
In the first step, 'step1_select_candidates', CharWidth and the CharHeight are used to select the candidates. The result of this step is also affected by ClutterSizeMax.
In the next step, 'step2_partition_characters', the parameter PartitionMethod and the parameter PartitionLines influence the result.
Step three, 'step3_connect_fragments', uses the parameters ConnectFragments and DotPrint. If dot-printed characters have to be detected and some dots are not connected to the character, there are two ways to overcome this problem: You can increase the FragmentDistance and/or decrease the StrokeWidth.
In the last step, 'step4_select_characters', the result is affected by the parameters DiacriticMarks and Punctuation.
DotPrint: Should be set to 'true' if dot prints should be read, else to 'false'.
StrokeWidth: Specifies the stroke width of the text. It is used to calculate internally used mask sizes to determine the characters. This mask sizes are also influenced through the parameters DotPrint, the average CharWidth, and the average CharHeight.
CharWidth: This can be a tuple with up to three values. The first value is the average width of a character. The second is the minimum width of a character and the third is the maximum width of a character. If the minimum is not set or equal -1, the operator automatically sets these value depending on average CharWidth. The same is the case if the maximum is not set. Some examples:
[10] sets the average character width to 10, the minimum and maximum are calculated by the operator.
[10,-1,20] sets the average character width to 10, the minimum value is calculated by the system, and the maximum is set to 20.
[10,5,20] sets the average character width to 10, the minimum to 5, and the maximum to 20.
CharHeight: This can be a tuple with up to three values. The first value is the average height of a character. The second is the minimum height of a character and the third is the maximum height of a character. If the minimum is not set or equal -1, the operator automatically sets these value depending on average CharHeight. The same is the case if the maximum is not set. Some examples:
[10] sets the average character height to 10, the minimum and maximum are calculated by the operator.
[10,-1,20] sets the average character height to 10 the minimum value is calculated by the system and the maximum is set to 20.
[10,5,20] this sets the average character height to 10, the minimum to 5 and the maximum to 20.
Punctuation: Set this parameter to 'true' if the operator also has to detect punctuation marks (e.g. .,:'`"), otherwise they will be suppressed.
DiacriticMarks: Set this parameter to 'true' if the text in your application contains diacritic marks (e.g. â,é,ö), or to 'false' to suppress them.
PartitionMethod: If neighboring characters are printed close to each other, they may be partly merged. With this parameter you can specify the method to partition such characters. The possible values are 'none', which means no partitioning is performed. 'fixed_width' means that the partitioning assumes a constant character width. If the width of the extracted region is well above the average CharWidth, the region ist split into parts that have the given average CharWidth. The partitioning starts at the left border of the region. 'variable_width' means that the characters are partitioned at the position where they have the thinnest connection. This method can be selected for characters that are printed with a variable-width font or if many consecutive characters are extracted as one symbol. It could be helpful to call text_line_slant and/or use text_line_orientation before calling select_characters.
PartitionLines: If some text lines or some characters of different text lines are connected, set this parameter to 'true'.
FragmentDistance: This parameter influences the connection of character fragments. If too much is connected, set the parameter to 'narrow' or 'medium'. In the case that more fragments should be connected, set the parameter to 'medium' or 'wide'. The connection is also influenced by the maximum of CharWidth and CharHeight. See also ConnectFragments.
ConnectFragments: Set this parameter to 'true' if the extracted symbols are fragmented, i.e., if a symbol is not extracted as one region but broken up into several parts. See also FragmentDistance and StopAfter in the step 'step3_connect_fragments'.
ClutterSizeMax: If the extracted characters contain clutter, i.e., small regions near the actual symbols, increase this value. If parts of the symbols are missing, decrease this value.
StopAfter: Use this parameter in the case the operator does not produce the desired results. By modifying this value the operator stops after the execution of the selected step and provides the corresponding results. To end on completion, set StopAfter to 'completion'.
Region of text lines in which to select the characters.
Selected characters.
Should dot print characters be detected?
Default value: 'false'
List of values: 'false', 'true'
Stroke width of a character.
Default value: 'medium'
List of values: 'bold', 'light', 'medium', 'ultra_light'
Width of a character.
Default value: 25
Typical range of values: 1 ≤ CharWidth
Restriction: CharWidth >= 1
Height of a character.
Default value: 25
Typical range of values: 1 ≤ CharHeight
Restriction: CharHeight >= 1
Add punctuation?
Default value: 'false'
List of values: 'false', 'true'
Exist diacritic marks?
Default value: 'false'
List of values: 'false', 'true'
Method to partition neighbored characters.
Default value: 'none'
List of values: 'fixed_width', 'none', 'variable_width'
Should lines be partitioned?
Default value: 'false'
List of values: 'false', 'true'
Distance of fragments.
Default value: 'medium'
List of values: 'medium', 'narrow', 'wide'
Connect fragments?
Default value: 'false'
List of values: 'false', 'true'
Maximum size of clutter.
Default value: 0
Typical range of values: 0 ≤ ClutterSizeMax
Restriction: 0 < ClutterSizeMax
Stop execution after this step.
Default value: 'completion'
List of values: 'completion', 'step1_select_candidates', 'step2_partition_characters', 'step3_connect_fragments', 'step4_select_characters'
for Index := 1 to 5 by 1
read_image (Image, 'dot_print_rotated/dot_print_rotated_'+Index$'02d')
text_line_orientation (Image, Image, 50, rad(-30), rad(30), OrientationAngle)
rotate_image (Image, ImageRotate, deg(-OrientationAngle), 'constant')
segment_characters (ImageRotate, ImageRotate, ImageForeground, \
RegionForeground, 'local_auto_shape', 'false', 'false', 'medium', \
25, 25, 0, 10, UsedThreshold)
select_characters (RegionForeground, RegionCharacters, 'true', 'ultra_light', \
[60,1,100], [60,1,100], 'false', 'false', 'none', 'true', 'wide', 'true', 0, \
'completion')
endfor
If the input parameters are set correctly, the operator select_characters returns the value 2 (H_MSG_TRUE). Otherwise an exception will be raised.
segment_characters, text_line_slant
do_ocr_single_class_mlp, do_ocr_multi_class_mlp
Foundation
| Operators |